Can you explain a bit more?
Sure. The idea in that Design by Playground piece is that designers can now build small, throwaway tools that answer one specific design question, instead of pushing pixels around in one general-purpose tool like Figma.
So what does that look like in practice? Picture a simple web page with your interface in the middle and a panel of controls on the side. You drag a slider and the text gets longer. You flip a toggle and the image goes missing. You change a timing value and the animation speeds up. You're manipulating variables and watching the real interface respond, instead of making a new mockup for every variation.
Here's an example from my own history of why that matters. When we redesigned Polar for iOS7, a lot of the hard problems only showed up in motion. We hid our headers on scroll, and iOS7's default transparency broke that behavior, so we ended up sliding the header under a thin blue underlay below the system bar. Then our pull to refresh elements started showing through the transparent header and made the text harder to read, so we blurred everything below it. None of that is visible in a static mockup. You have to scroll, pull, and watch it happen on a real device.
Same with visual hierarchy. In a flat design, a primary action like Agree might be just a bit bigger and a bit bolder than everything else around it, and getting that right with less contrast to work with is all about iteration. A playground with sliders for size, weight, and color lets you run through dozens of those iterations in minutes instead of redrawing each one.
So why didn't we all work this way years ago? Building a custom tool used to take longer than just fiddling with mockups, especially for something you'd use once. AI coding agents changed that math. Now you can spin up a playground in minutes, run it in the actual environment your product ships in (real fonts, real scrolling, real touch behavior), and export the values you settle on as code or design tokens.
That matters even more for new territory like AI agent interfaces. Our first inclination with any new medium is to copy what worked before, and the first things we try are very likely going to be the wrong things. A playground lets you try a lot of wrong things quickly and cheaply until you find what works.
My practical advice: next time you've got one nagging design question, ask a coding agent to build you a tool that answers just that question. Use it for an afternoon, then toss it.